Research on FPGA Implementation of a Fire Detection System for Key Equipment in Converter Stations

抄録

An FPGA-based fire detection system using a back propagation (BP) neural network was designed for early fire detection in key equipment in converter stations. An 8-5-1 BP network structure was trained, achieving a recognition accuracy of 94.08%. Fixed-point data quantization and pipelining were employed to reduce computational complexity, lowering resource consumption and enhancing speed. The FPGA system used 683 LUTs, achieved a 94.6% detection rate, consumed only 1.342 W of power and completed a single detection in 3.25 μs,a significant improvement compared to the 8.56 ms detection time on MATLAB.This system demonstrates excellent reliability, real-time performance, and promising application potential for early fire detection in key equipment in converter stations.